Re: [music-dsp] Dither video and articles
Nigel, I looked at your video again and it seems to me it's confusing as to whether you mean 'don't dither the 24b final output' or 'don't ever dither at 24b'. You make statements several times that imply the former, but in your discussion about 24b on all digital interfaces, sends and receives etc, you clearly say to never dither at 24b. Several people in this thread have pointed out the difference between intermediate stage truncation and final stage truncation, and the fact that if truncation is done repeatedly, any distortion spectra will continue to build. It is not noise-like, the peaks are coherent peaks and correlated to the signal. You don't say in the video what the processing history is for the files you are using. If they are simple captures with no processing, they probably reflect the additive gaussian noise present at the 20th bit in the A/D, based on Andy's post, and are properly dithered for 24b truncation. My point is that at the digital capture stage you have (S+N) and the amplitude distribution of the S+N signal might be fine for 24b truncation if N is dither-like. After various stages of digital processing including non-linear steps, the (S+N) intermediate signal may no longer have an adequate amplitude distribution to be truncated without 24b dither. I think the whole subject of self dither might be better approached through FFT measurement than by listening. Bob Katz shows an FFT of truncation spectra at 24b in his book on 'Itunes Music, Mastering for High Resolution Audio Delivery' but he uses a generated, dithered pure tone that doesn't start with added gaussian noise. Haven't thought about it but I can imagine extending his approach into a research effort. Offhand I don't know anything that would go wrong in your difference file ( ...if the error doesn't sound wrong). It's a common method for looking at residuals. Vicki On Feb 8, 2015, at 6:11 PM, Nigel Redmon wrote: Beyond that, Nigel raises this issue in the context of self-dither”... First, remember that I’m the guy who recommended “always” dithering 16-bit (no “always” as in “alway necessary”, but as in “do it always, unless you know that it gives no improvement”), and to not bother dithering 24-bit. So, I’m only interested in this discussion for 24-bit. That said: ...In situations where there is a clear external noise source present, whether the situation is analog to digital conversion or digital to digital bit depth change, the external noise may, or may not, be satisfactory as dither but at least it's properties can be measured. For 24-bit audio, could you give an example of when it’s likely to not be satisfactory (maybe you’ve already given a reference to determining “satisfactory)? Offhand, I’d say one case might be with extremely low noise, then digitally faded such that you fade the noise level below the dithering threshold while you still have enough signal to exhibit truncation distortion, and the fade characteristics allow it to last long enough to matter to your ears—if we weren’t talking about this distortion being down near -140 dB in the first place. I’d think that, typically, you’d have gaussian noise at a much higher level that is needed to dither 24-bit; that could change with digital processing, but I think that in the usual recording chain, it seems pretty hard to avoid for your analog to digital conversion” case. I’m still interested in what you have to say about my post yesterday (“...if the error doesn’t sound wrong to the ear, can it still sound wrong added to the music?”). Care to comment? On Feb 8, 2015, at 8:09 AM, Vicki Melchior vmelch...@earthlink.net wrote: I have no argument at all with the cheap high-pass TPDF dither; whenever it was published the original authors undoubtedly verified that the moment decoupling occurred, as you say. And that's what is needed for dither effectiveness. If you're creating noise for dither, you have the option to verify its properties. But in the situation of an analog signal with added, independent instrument noise, you do need to verify that the composite noise source actually satisfies the criteria for dither. 1/f noise in particular has been questioned, which is why I raised the spectrum issue. Beyond that, Nigel raises this issue in the context of self-dither. In situations where there is a clear external noise source present, whether the situation is analog to digital conversion or digital to digital bit depth change, the external noise may, or may not, be satisfactory as dither but at least it's properties can be measured. If the 'self-dithering' instead refers to analog noise captured into the digitized signal with the idea that this noise is going to be preserved and available at later truncation steps to 'self dither' it is a very very hazy argument. I'm aware of the various caveats that are often
Re: [music-dsp] Dither video and articles
That's a clear explanation of the self-dither assumed in A/D conversion, thanks for posting it. Vicki On Feb 8, 2015, at 9:11 PM, Andrew Simper wrote: Vicki, If you look at the limits of what is possible in a real world ADC there is a certain amount of noise in any electrical system due to gaussian thermal noise: http://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise For example if you look at an instrument / measurement grade ADC like this: http://www.prismsound.com/test_measure/products_subs/dscope/dscope_spec.php They publish figures of a residual noise floor of 1.4 uV, which they say is -115 dBu. So if you digitise a 1 V peak (2 V peak to peak) sine wave with a 24-bit ADC then you will have hiss (which includes a large portion of gaussian noise) at around the 20 bit mark, so you will have 4-bits of hiss to self dither. This has nothing to do with microphones or noise in air, this is in the near perfect case of transmission via a well shielded differential cable transferring the voltage directly to the ADC. All the best, Andy -- cytomic -- sound music software -- On 9 February 2015 at 00:09, Vicki Melchior vmelch...@earthlink.net wrote: I have no argument at all with the cheap high-pass TPDF dither; whenever it was published the original authors undoubtedly verified that the moment decoupling occurred, as you say. And that's what is needed for dither effectiveness. If you're creating noise for dither, you have the option to verify its properties. But in the situation of an analog signal with added, independent instrument noise, you do need to verify that the composite noise source actually satisfies the criteria for dither. 1/f noise in particular has been questioned, which is why I raised the spectrum issue. Beyond that, Nigel raises this issue in the context of self-dither. In situations where there is a clear external noise source present, whether the situation is analog to digital conversion or digital to digital bit depth change, the external noise may, or may not, be satisfactory as dither but at least it's properties can be measured. If the 'self-dithering' instead refers to analog noise captured into the digitized signal with the idea that this noise is going to be preserved and available at later truncation steps to 'self dither' it is a very very hazy argument. I'm aware of the various caveats that are often postulated, i.e. signal is captured at double precision, no truncation, very selected processing. But even in minimalist recording such as live to two track, it's not clear to me that the signal can get through the digital stages of the A/D and still retain an unaltered noise distribution. It certainly won't do so after considerable processing. So the sho r t answer is, dither! At the 24th bit or at the 16th bit, whatever your output is. If you (Nigel or RBJ) have references to the contrary, please say so. Vicki On Feb 8, 2015, at 10:11 AM, robert bristow-johnson wrote: On 2/7/15 8:54 AM, Vicki Melchior wrote: Well, the point of dither is to reduce correlation between the signal and quantization noise. Its effectiveness requires that the error signal has given properties; the mean error should be zero and the RMS error should be independent of the signal. The best known examples satisfying those conditions are white Gaussian noise at ~ 6dB above the RMS quantization level and white TPDF noise at ~3dB above the same, with Gaussian noise eliminating correlation entirely and TPDF dither eliminating correlation with the first two moments of the error distribution. That's all textbook stuff. There are certainly noise shaping algorithms that shape either the sum of white dither and quantization noise or the white dither and quantization noise independently, and even (to my knowledge) a few completely non-white dithers that are known to work, but determining the effectiveness of noise at dithering still requires examining the statistical properties of the error signal and showi n g th at the mean is 0 and the second moment is signal independent. (I think Stanley Lipschitz showed that the higher moments don't matter to audibility.) but my question was not about the p.d.f. of the dither (to decouple both the mean and the variance of the quantization error, you need triangular p.d.f. dither of 2 LSBs width that is independent of the *signal*) but about the spectrum of the dither. and Nigel mentioned this already, but you can cheaply make high-pass TPDF dither with a single (decent) uniform p.d.f. random number per sample and running that through a simple 1st-order FIR which has +1 an -1 coefficients (i.e. subtract the previous UPDF from the current UPDF to get the high-pass TPDF). also, i think Bart Locanthi (is he still on this planet?) and someone else did a simple paper back in the 90s about the possible benefits of
Re: [music-dsp] Dither video and articles
But it matters, because the whole point of dithering to 16bit depends on how common that ability is. Depends on how common? I’m not sure what qualifies for common, but if it’s 1 in 100, or 5 in 100, it’s still a no-brainer because it costs nothing, effectively. But more importantly, I don’t think you’re impressed by my point that it’s the audio engineers, the folks making the music, that are in the best position to hear it, and to do something about it. There are the ones listening carefully, in studios built to be quiet and lack reflections and resonances that might mask things, on revealing monitors and with ample power. I don’t think that you understand that it’s these guys who are not going to let their work go out the door with grit on it, even if it’s below -90 dB. You wouldn’t get many sympathetic ears among them if you advocated that they cease this dithering nonsense :-) I get enough grief about telling them that dither at 24-bit is useless. How common it is for for the average listener is immaterial. It’s not done for the average listener. On Feb 9, 2015, at 6:56 PM, Didier Dambrin di...@skynet.be wrote: I'm having a hard time finding anyone who could hear past the -72dB noise, here around. Really, either you have super-ears, or the cause is (technically) somewhere else. But it matters, because the whole point of dithering to 16bit depends on how common that ability is. -Message d'origine- From: Andrew Simper Sent: Saturday, February 07, 2015 2:08 PM To: A discussion list for music-related DSP Subject: Re: [music-dsp] Dither video and articles On 7 February 2015 at 03:52, Didier Dambrin di...@skynet.be wrote: It was just several times the same fading in/out noise at different levels, just to see if you hear quieter things than I do, I thought you'd have guessed that. https://drive.google.com/file/d/0B6Cr7wjQ2EPub2I1aGExVmJCNzA/view?usp=sharing (0dB, -36dB, -54dB, -66dB, -72dB, -78dB) Here if I make the starting noise annoying, then I hear the first 4 parts, until 18:00. Thus, if 0dB is my threshold of annoyance, I can't hear -72dB. So you hear it at -78dB? Would be interesting to know how many can, and if it's subjective or a matter of testing environment (the variable already being the 0dB annoyance starting point) Yep, I could hear all of them, and the time I couldn't hear the hiss any more as at the 28.7 second mark, just before the end of the file. For reference this noise blast sounded much louder than the bass tone that Nigel posted when both were normalised, I had my headphones amp at -18 dB so the first noise peak was loud but not uncomfortable. I thought it was an odd test since the test file just stopped before I couldn't hear the LFO amplitude modulation cycles, so I wasn't sure what you were trying to prove! All the best, Andy -Message d'origine- From: Andrew Simper Sent: Friday, February 06, 2015 3:21 PM To: A discussion list for music-related DSP Subject: Re: [music-dsp] Dither video and articles Sorry, you said until, which is even more confusing. There are multiple points when I hear the noise until since it sounds like the noise is modulated in amplitude by a sine like LFO for the entire file, so the volume of the noise ramps up and down in a cyclic manner. The last ramping I hear fades out at around the 28.7 second mark when it is hard to tell if it just ramps out at that point or is just on the verge of ramping up again and then the file ends at 28.93 seconds. I have not tried to measure the LFO wavelength or any other such things, this is just going on listening alone. All the best, Andrew Simper On 6 February 2015 at 22:01, Andrew Simper a...@cytomic.com wrote: On 6 February 2015 at 17:32, Didier Dambrin di...@skynet.be wrote: Just out of curiosity, until which point do you hear the noise in this little test (a 32bit float wav), starting from a bearable first part? https://drive.google.com/file/d/0B6Cr7wjQ2EPucjFCSUhGNkVRaUE/view?usp=sharing I hear noise immediately in that recording, it's hard to tell exactly the time I can first hear it since there is some latency from when I press play to when the sound starts, but as far as I can tell it is straight away. Why do you ask such silly questions? All the best, Andrew Simper -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp - Aucun virus trouve dans ce message. Analyse effectuee par AVG - www.avg.fr Version: 2015.0.5645 / Base de donnees virale: 4281/9068 - Date: 06/02/2015 -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp
Re: [music-dsp] Dither video and articles
I'm having a hard time finding anyone who could hear past the -72dB noise, here around. Really, either you have super-ears, or the cause is (technically) somewhere else. But it matters, because the whole point of dithering to 16bit depends on how common that ability is. -Message d'origine- From: Andrew Simper Sent: Saturday, February 07, 2015 2:08 PM To: A discussion list for music-related DSP Subject: Re: [music-dsp] Dither video and articles On 7 February 2015 at 03:52, Didier Dambrin di...@skynet.be wrote: It was just several times the same fading in/out noise at different levels, just to see if you hear quieter things than I do, I thought you'd have guessed that. https://drive.google.com/file/d/0B6Cr7wjQ2EPub2I1aGExVmJCNzA/view?usp=sharing (0dB, -36dB, -54dB, -66dB, -72dB, -78dB) Here if I make the starting noise annoying, then I hear the first 4 parts, until 18:00. Thus, if 0dB is my threshold of annoyance, I can't hear -72dB. So you hear it at -78dB? Would be interesting to know how many can, and if it's subjective or a matter of testing environment (the variable already being the 0dB annoyance starting point) Yep, I could hear all of them, and the time I couldn't hear the hiss any more as at the 28.7 second mark, just before the end of the file. For reference this noise blast sounded much louder than the bass tone that Nigel posted when both were normalised, I had my headphones amp at -18 dB so the first noise peak was loud but not uncomfortable. I thought it was an odd test since the test file just stopped before I couldn't hear the LFO amplitude modulation cycles, so I wasn't sure what you were trying to prove! All the best, Andy -Message d'origine- From: Andrew Simper Sent: Friday, February 06, 2015 3:21 PM To: A discussion list for music-related DSP Subject: Re: [music-dsp] Dither video and articles Sorry, you said until, which is even more confusing. There are multiple points when I hear the noise until since it sounds like the noise is modulated in amplitude by a sine like LFO for the entire file, so the volume of the noise ramps up and down in a cyclic manner. The last ramping I hear fades out at around the 28.7 second mark when it is hard to tell if it just ramps out at that point or is just on the verge of ramping up again and then the file ends at 28.93 seconds. I have not tried to measure the LFO wavelength or any other such things, this is just going on listening alone. All the best, Andrew Simper On 6 February 2015 at 22:01, Andrew Simper a...@cytomic.com wrote: On 6 February 2015 at 17:32, Didier Dambrin di...@skynet.be wrote: Just out of curiosity, until which point do you hear the noise in this little test (a 32bit float wav), starting from a bearable first part? https://drive.google.com/file/d/0B6Cr7wjQ2EPucjFCSUhGNkVRaUE/view?usp=sharing I hear noise immediately in that recording, it's hard to tell exactly the time I can first hear it since there is some latency from when I press play to when the sound starts, but as far as I can tell it is straight away. Why do you ask such silly questions? All the best, Andrew Simper -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp - Aucun virus trouve dans ce message. Analyse effectuee par AVG - www.avg.fr Version: 2015.0.5645 / Base de donnees virale: 4281/9068 - Date: 06/02/2015 -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp - Aucun virus trouve dans ce message. Analyse effectuee par AVG - www.avg.fr Version: 2015.0.5645 / Base de donnees virale: 4281/9071 - Date: 07/02/2015 -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Dither video and articles
OK, I don’t want to diverge too much from the practical to the theoretical, so I’m going to run down what is usual, not what is possible, because it narrows the field of discussion. Most people I know are using recording systems that bussing audio at 32-bit float, minimum, and use 64-bit float calculations in plug-ins and significant processing. They may still be using 24-bit audio tracks on disk, but for the most part they are recorded and are dithered one way or another (primarily gaussian noise in the recording process). They may bounce things to tracks to free processor cycles. I think in large majority of cases, these are self-dithered, but even if it doesn’t happen for some, I don’t think it will impact the audio. And if people are worried about it, I don’t understand why they aren’t using 32-bit float files, as I think most people have that choices these days. Some of the more hard core will send audio out to a convertor (therefore truncated at 24-bit), and back in. Again, I think the vast majority of cases, these will self dither, but then there’s the fact error is at a very low level, will get buried in the thermal noise of the electronics, etc. Maybe I left out some other good ones, but to cut it short, yes, I’m mainly talking about final mixes. At 24-bit, that often goes to someone else to master. The funny thing is that some mastering engineers say “only dither once!”, and they want to be the one doing it. Others point out that they may want to mess with the dynamic range and boost frequencies, and any error from not dithering 24-bit will show up in…you know, the stereo imaging, depth, etc. I think it would be exceptional to actually have truncation distortion of significant duration, except for potential situations with unusual fades, so I’m not worried about saying don’t dither 24-bit, even heading to a mastering engineer (but again, do it if you want, it’s just no big deal for final outputs–in contrast to the pain in the rear it is to do it at every point for the items I mentioned in previous paragraphs). Down the more theoretical paths, I’ve had people argue that this is a big deal because things like ProTools 56k plug-ins need to be dithered internally…but why argue legacy stuff that “is what it is”, and secondly, these people usually don’t think through how many 24-bit truncations occur in a 56k algorithm, and you only have so many cycles. The other thing I sometimes get is the specter of the cumulative effect (but what if you have so many tracks, and feedback, and…)—but it seems to me that the more of this you get going on, to approach a meaningful error magnitude, the more it’s jumbled up in chaos and the less easy it is for your ear to recognize it as “bad”. On Feb 9, 2015, at 7:54 AM, Vicki Melchior vmelch...@earthlink.net wrote: Nigel, I looked at your video again and it seems to me it's confusing as to whether you mean 'don't dither the 24b final output' or 'don't ever dither at 24b'. You make statements several times that imply the former, but in your discussion about 24b on all digital interfaces, sends and receives etc, you clearly say to never dither at 24b. Several people in this thread have pointed out the difference between intermediate stage truncation and final stage truncation, and the fact that if truncation is done repeatedly, any distortion spectra will continue to build. It is not noise-like, the peaks are coherent peaks and correlated to the signal. You don't say in the video what the processing history is for the files you are using. If they are simple captures with no processing, they probably reflect the additive gaussian noise present at the 20th bit in the A/D, based on Andy's post, and are properly dithered for 24b truncation. My point is that at the digital capture stage you have (S+N) and the amplitude distribution of the S+N signal might be fine for 24b truncation if N is dither-like. After various stages of digital processing including non-linear steps, the (S+N) intermediate signal may no longer have an adequate amplitude distribution to be truncated without 24b dither. I think the whole subject of self dither might be better approached through FFT measurement than by listening. Bob Katz shows an FFT of truncation spectra at 24b in his book on 'Itunes Music, Mastering for High Resolution Audio Delivery' but he uses a generated, dithered pure tone that doesn't start with added gaussian noise. Haven't thought about it but I can imagine extending his approach into a research effort. Offhand I don't know anything that would go wrong in your difference file ( ...if the error doesn't sound wrong). It's a common method for looking at residuals. Vicki On Feb 8, 2015, at 6:11 PM, Nigel Redmon wrote: Beyond that, Nigel raises this issue in the context of self-dither”... First, remember that I’m the guy who recommended “always”
Re: [music-dsp] Dither video and articles
I’m thankful for Andy posting that clear explanation too. Sometimes I understate things—when I said that it would be “pretty hard to avoid” having ample gaussian noise to self-dither in the A/D process, I was thinking cryogenics (LOL). On Feb 9, 2015, at 7:54 AM, Vicki Melchior vmelch...@earthlink.net wrote: That's a clear explanation of the self-dither assumed in A/D conversion, thanks for posting it. Vicki On Feb 8, 2015, at 9:11 PM, Andrew Simper wrote: Vicki, If you look at the limits of what is possible in a real world ADC there is a certain amount of noise in any electrical system due to gaussian thermal noise: http://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise For example if you look at an instrument / measurement grade ADC like this: http://www.prismsound.com/test_measure/products_subs/dscope/dscope_spec.php They publish figures of a residual noise floor of 1.4 uV, which they say is -115 dBu. So if you digitise a 1 V peak (2 V peak to peak) sine wave with a 24-bit ADC then you will have hiss (which includes a large portion of gaussian noise) at around the 20 bit mark, so you will have 4-bits of hiss to self dither. This has nothing to do with microphones or noise in air, this is in the near perfect case of transmission via a well shielded differential cable transferring the voltage directly to the ADC. All the best, Andy -- cytomic -- sound music software -- On 9 February 2015 at 00:09, Vicki Melchior vmelch...@earthlink.net wrote: I have no argument at all with the cheap high-pass TPDF dither; whenever it was published the original authors undoubtedly verified that the moment decoupling occurred, as you say. And that's what is needed for dither effectiveness. If you're creating noise for dither, you have the option to verify its properties. But in the situation of an analog signal with added, independent instrument noise, you do need to verify that the composite noise source actually satisfies the criteria for dither. 1/f noise in particular has been questioned, which is why I raised the spectrum issue. Beyond that, Nigel raises this issue in the context of self-dither. In situations where there is a clear external noise source present, whether the situation is analog to digital conversion or digital to digital bit depth change, the external noise may, or may not, be satisfactory as dither but at least it's properties can be measured. If the 'self-dithering' instead refers to analog noise captured into the digitized signal with the idea that this noise is going to be preserved and available at later truncation steps to 'self dither' it is a very very hazy argument. I'm aware of the various caveats that are often postulated, i.e. signal is captured at double precision, no truncation, very selected processing. But even in minimalist recording such as live to two track, it's not clear to me that the signal can get through the digital stages of the A/D and still retain an unaltered noise distribution. It certainly won't do so after considerable processing. So the sho r t answer is, dither! At the 24th bit or at the 16th bit, whatever your output is. If you (Nigel or RBJ) have references to the contrary, please say so. Vicki On Feb 8, 2015, at 10:11 AM, robert bristow-johnson wrote: On 2/7/15 8:54 AM, Vicki Melchior wrote: Well, the point of dither is to reduce correlation between the signal and quantization noise. Its effectiveness requires that the error signal has given properties; the mean error should be zero and the RMS error should be independent of the signal. The best known examples satisfying those conditions are white Gaussian noise at ~ 6dB above the RMS quantization level and white TPDF noise at ~3dB above the same, with Gaussian noise eliminating correlation entirely and TPDF dither eliminating correlation with the first two moments of the error distribution. That's all textbook stuff. There are certainly noise shaping algorithms that shape either the sum of white dither and quantization noise or the white dither and quantization noise independently, and even (to my knowledge) a few completely non-white dithers that are known to work, but determining the effectiveness of noise at dithering still requires examining the statistical properties of the error signal and showi n g th at the mean is 0 and the second moment is signal independent. (I think Stanley Lipschitz showed that the higher moments don't matter to audibility.) but my question was not about the p.d.f. of the dither (to decouple both the mean and the variance of the quantization error, you need triangular p.d.f. dither of 2 LSBs width that is independent of the *signal*) but about the spectrum of the dither. and Nigel mentioned this already, but you can cheaply make high-pass TPDF dither with a single (decent) uniform p.d.f. random
